Federated learning (FL) for the Internet of Things (IoT) has become a rapidly growing research area, addressing the need for distributed intelligence while preserving data privacy across resource-constrained devices. Research papers in this domain focus on enabling collaborative machine learning without centralizing sensitive IoT data, thus ensuring security and compliance in applications such as healthcare, smart homes, industrial automation, and autonomous vehicles. Key contributions explore lightweight FL algorithms, communication-efficient aggregation, model personalization, and energy-aware training tailored for heterogeneous IoT devices. Challenges such as data heterogeneity, straggler devices, adversarial attacks, and limited bandwidth are actively studied, with solutions leveraging edge computing, blockchain, differential privacy, and secure multiparty computation. Recent works also emphasize the integration of FL with 5G/6G networks, digital twins, and zero trust architectures to enhance scalability, trustworthiness, and real-time decision-making in IoT ecosystems.